Optimization of L-asparaginase production by Aspergillus terreus MTCC 1782 using response surface methodology and artificial neural network-linked genetic algorithm

The sequential optimization strategy of design of experiments and back propagation algorithm of artificial neural network-linked genetic algorithm were used to find the significant fermentation media components and optimum concentration for maximum L-asparaginase production by Aspergillus terreus MTCC 1782 in submerged fermentation. Components such as L-proline, sodium nitrate, L-asparagine and glucose were identified as significant fermentation media components using PlackettBurman design. The central composite design was used to fit the second-order polynomial model describing the effect of significant media components on L-asparaginase production with coefficient of determination (R2) 0.973. A nonlinear model was developed with high coefficient of determination (R2) 0.997 using incremental back propagation algorithm of neural network. The high value of coefficient of determination for artificial neural network model justified an excellent correlation between variables and L-asparaginase activity and found to be more efficient than the second-order polynomial model of central composite design. Hence, the optimum concentration of the significant media components was determined using artificial neural network-linked genetic algorithm. The predicted optimum concentration of the media components was L-proline 1.7% (w/v), sodium nitrate 1.99% (w/v), L-asparagine 1.38% (w/v) and glucose 0.65% (w/v) with an experimentally confirmed L-asparaginase activity of 40.86 IU mL-1. Copyright (C) 2010 Curtin University of Technology and John Wiley & Sons, Ltd.